Client discovery — raw material before copy

Killer copy is not born from imagination: it is born from the audience's real phrases, real objections and a real differentiator. This skill turns the 5-line brief from ABOUT.md into docs/discovery.md — the mandatory input for the copywriter and the landing blueprint.

Timebox: 60–90 min of research. Endless discovery is procrastination dressed up as diligence. Collect, synthesize, move on.

Step 0 — Inputs

1. Read ABOUT.md: client, segment, audience, primary conversion, tone.
2. Ask the user (AskUserQuestion, one round only):
   • Who are the top 3 competitors? (if unknown, you find them in Step 2)
   • Is there existing client material? (testimonials, surveys, sales-call recordings, current site)
   • What does the client THINK their differentiator is?

Step 1 — Customer voice (the heart of the skill)

Mine LITERAL phrases from the audience. Sources, via WebSearch/WebFetch:

• The client's own reviews: Google Maps, Trustpilot, G2, app stores.
• Competitors' reviews (their unmet pains are your argument).
• Reddit, forums, YouTube comments in the niche — search "[niche] worth it", "[niche] experience", "how to solve [problem]".
• Market-specific sources: see .claude/locales/<language>.md when the project Language has one — it lists the review platforms and channels that matter in that market.

What to extract from each source (with literal quote + link):

• Pains — what they complain about, in their exact words ("slow to reply", "hidden pricing").
• Desires — the dream outcome, as they describe it.
• Objections — why they hesitate before buying ("will this work for my case?").
• Language — what they call things (does the audience say "consultation" or "assessment"? "site" or "page"?).

Golden rule: literal quote > paraphrase. The headline that converts has almost always already been written by a customer in a review.

Client with no online presence (new business)? Replace with a guided interview with the user: 5 questions — who is the ideal customer who would buy today; what have they tried before; what would make them say no; what do competitors promise; what concrete result in 90 days.

Step 2 — Competitors (3 to 5)

For each one, open the landing/site (Playwright if available, otherwise WebFetch) and record:

Step 3 — Defensible differentiator

Cross-reference: what the client has × what the audience wants × what competitors don't deliver. A differentiator only counts if it passes the test: a competitor could NOT say the same sentence tomorrow. "Quality and service" fails; "48-hour delivery with tracking" passes.

If no real differentiator exists, SAY SO in the document: the strategy becomes competing on clarity, proof and response speed — never inventing a fake differentiator.

Output — docs/discovery.md (fixed format)

# Discovery — [client] ([date])
## Audience
Who they are + how they talk (3-5 lines, with the niche-term glossary)
## Pains (top 5, with literal quotes and source)
## Desires / dream outcome (with quotes)
## Objections (top 5) + honest answer to each
## Competitors (Step 2 table)
## Defensible differentiator (or "parity — compete on clarity and proof")
## Candidate copy angles (3, one paragraph each)

Anti-patterns

• Inventing a pain or testimonial the research didn't find — the document only contains what has a source.
• Paraphrasing everything into "marketing language" — kills exactly the value of the research.
• Copying the strongest competitor's promise (if they already own it, you arrive second).
• Researching for 4 hours — timebox blown = synthesize with what you have and note the gaps.